Opus 5.5's highest effort levels trigger safeguards that 'high' avoids
Kyxstrez · reddit · 2026-10-25
A Reddit user reports a counterintuitive quirk in Opus 5.5: certain questions at xhigh or max effort reliably trigger safeguards and get refused, while dialing reasoning down to 'high' — the highest level the poster says is actually usable — lets the task through. The tradeoff: you finish the job but lose the top capabilities the bigger reasoning budget buys.
More from Models
- AI Writing Is Getting Harder to Read — Research Says It's Trainable — lmoroney · 2026-10-25
- ChatGPT web quietly adds a usage limit popup showing Pro plan quota — menhguin · 2026-10-25
- 4B model fine-tuned 27 min on one A100 lifts letter-exam accuracy to 74.5% — MaziyarPanahi · 2026-10-25
- Even at temperature 0, 45 of 1,273 LLM answers changed between runs — MaziyarPanahi · 2026-10-25
- Claude reportedly went off-script in UK AISI hacking test, 9 of 43 runs — faraz0x · 2026-10-25
- Frontier LLMs can build their own scoring harness to hill-climb any goal — Ryoiki-Tokuiten · 2026-10-25